I have samples from RNAsequencing that I would like advice on the best way to identify gene expression profiles.
I have 2 genotypes and 2 different temperature conditions: Control wildtype at 28C, mutant at 28C, wildtype at 36C and mutant at 36C. Ideally I would like to be able to identify genes in different expression patterns across these groups. For example, genes that are only up-regulated in mutant at 28C, or genes that are up in both genotypes at 28C and down in 36C genotypes. Is it enough to do differential expression comparing all conditions to control wildtype at 28C, and filter out genes based on that expression? Or is there a better way?
Thank you!
1 answer
You want the interaction term, not three contrasts against WT 28C. It's a clean 2x2, so ~ genotype * temperature, and the interaction is basically "does the temperature response depend on genotype", which is most of what you're describing.
Main reason to avoid the filter-the-contrasts route: if a gene is significant in mutant vs WT at 28 and not at 36, that isn't evidence the two differ. Significant versus non-significant isn't itself a comparison. Easy to lose a lot of time to that one.
For the pattern side, an LRT against a reduced model gets you "something is happening across the four groups", then cluster those genes on their scaled group means. degPatterns in DEGreport does it out of the box if you'd rather not hand-roll it.
Log in to answer this question.
Do you have replicates? If you know how to use R the edgeR userguide has some great discussion of your problem: https://www.bioconductor.org/packages/release/bioc/vignettes/edgeR/inst/doc/edgeRUsersGuide.pdf . What platform do you prefer to use for this kind of analysis?